Estimation of event based runoff coefficient using artificial intelligence models (Case study: Kasilian watershed)

In this research, estimation of the Runoff Coefficient (RC) is carried out depending on land cover. Initially, RC modeling was performed using 54 hourly rainfall and corresponding runoff data during the period 1987–2010 in the Kasilian watershed. Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS) and Support Vector Regression (SVR) models and effective factors including rainfall intensity, Φ index (the average loss), five-day previous rainfall and Normalized Difference Vegetation Index (NDVI) were used to estimate RC. The results showed that the ANN model was more efficient than the other two models and had Mean Bias Error (MBE), Coefficient of Determination (R2), Nash–Sutcliffe Efficiency (NSE) and Normalized Root Mean Square Error (NRMSE) equal to 0.08, 0.85, 0.84 and 0.37, respectively for the training phase and 0.12, 0.76, 0.74 and 0.47 for the test phase. In general, it is suggested that RC plays a major role in hydrological mechanisms and flooding. Thus, optimal estimation of RC can be helpful in better management of soil and water conservation and erosion and sediment management in this area.